1. Field of the Invention
The present invention relates to a preprocessing apparatus connected to an apparatus for diagnosing a psychosomatic state of a voice uttered by a human being, an apparatus for diagnosing the brain function, an apparatus for diagnosing a human factor, an apparatus for diagnosing affinity, an apparatus for diagnosing a vocational aptitude, and soon. The present invention belongs to the field of the medical diagnosis technique.
2. Background of the Related Art
In a conventional chaos theoretical diagnosis system or apparatus using uttered voice, a numerical value is used as a diagnosis subject data. This numeric value can be obtained by mechanically dividing a continuous uttered voice into specific processing unit time intervals, such as voice signals each having a temporal length of one second, calculating the first Lyapunov exponent in each processing unit time interval, and conducting statistical processing such as moving average calculation on the calculated first Lyapunov exponent.
If voice data of the processing unit time interval to be used in the chaos theoretical diagnosis is mechanically cut out from an input voice, then a subtle difference in a cutout point in time exerts influence upon the calculated first Lyapunov exponent and causes variation in the diagnosis value.
For reducing the variation in the diagnosis value, it is necessary to set a time that is sufficiently long as compared with the cut out time. For example, by setting an average time in the range of several tens of seconds to several minutes in the case where the processing unit time is set equal to one second. It is also necessary to calculate a temporal average value from the first Lyapunov exponent calculated in respective processing units, and use the temporal average value as a diagnosis value.
When making a chaos theoretical diagnosis from the uttered voice, for example, a continuous uttered voice ranging over at least several minutes is needed as input data in order to get a high diagnosis precision, such as to reduce an error in diagnosis value to several percents or less.
In addition, in the conventional voice signal processing method, it has been impossible to make a significant diagnosis by using a voice signal taken for an interval of approximately several times as long as the processing unit time, such as a voice signal taken for several seconds in the case where the processing unit time interval is one second.
Even if the processing unit time is made shorter, merely the number of the first Lyapunov exponent that must be calculated increases. For example, if the processing unit time is shortened from one second to 0.1 second, the number of the first Lyapunov exponent that must be calculated increases to ten times. Even if a time interval as long as ten times or more is spent for computation processing, the variation in Lyapunov exponent caused by deviation in a point in time, at which each diagnosis data is cutout, cannot be effectively reduced. It is impossible to make a significant diagnosis by using a voice taken for an interval of several seconds.
In the conventional technique described above, it has become an important problem to clarify a method for reducing the variation in the calculated first Lyapunov exponent caused by a difference in a point of time at which diagnosis data is cut out from voice data used in the chaos theoretical diagnosis. Furthermore, a method for calculating a comparatively highly reliable diagnosis value from comparatively short voice data has become an important problem in the same way.
The present invention has been developed to solve the above-described problems. An object of the present invention is to provide an apparatus that reduces the variation in the first Lyapunov exponent calculated using a chaos theoretical diagnosis by cutting out unit voice data used in the chaos theoretical diagnosis while taking a phoneme as the unit, and that calculates a diagnosis value having a reliability equal to or higher than that obtained by using the conventional technique, on the basis of voice data shorter than that in the case using the conventional technique.